AI Accused of Destroying Millions of Books; 12+ Groups Urge FTC

For AI practitioners and model developers, the letter underscores a hard truth: high-quality copyrighted books remain irreplaceable long-form training data, but how they are acquired and disposed of is becoming a compliance liability. If FTC scrutiny grows, engineering teams may need to document provenance, preservation, and access protocols for every corpus, not just licensing. The hoard-and-destroy allegation also raises a strategic question about whether destroying rare source material deliberately excludes competitors and centralizes model quality among the few incumbents who can afford to hoard it.

On August 21, 2026, a coalition of more than a dozen public interest and consumer advocacy organizations asked the Federal Trade Commission to investigate whether leading AI developers are buying, scanning, and then destroying millions of physical books to build large language models. The letter, sent Friday to FTC Chairman Andrew Ferguson and Commissioner Mark Meador, alleges an intentional "hoard-and-destroy" strategy. Signatories including Demand Progress Education Fund, Consumer Federation of America, and Institute for Local Self-Reliance contend that some destroyed volumes may be among the last surviving physical copies of their texts. They ask the FTC to determine how widespread the practice is and how many discarded books were the final known copies of the works they contained.

“

Signatories including Demand Progress Education Fund, Consumer Federation of America, and Institute for Local Self-Reliance contend that some destroyed volumes may be among the last surviving physical copies of their texts.

The letter links the alleged practice to Section 5 of the FTC Act, which prohibits unfair methods of competition and unfair or deceptive acts or practices in commerce. The groups describe a future "where only the wealthiest incumbents can build high-quality AI models and operate as the sole holders of humanity's written works — after having destroyed the originals to get there." That positions book destruction not merely as cultural loss but as a market-structure issue: if rare physical texts are removed, researchers, libraries, and rival AI developers may lose access to nonrenewable source material, reinforcing the data advantage of well-capitalized incumbents.

Large language models rely on vast, diverse text corpora, and edited long-form books are especially valuable for generating coherent, knowledgeable outputs. As web text and social data face quality or legal constraints, book scanning has become an attractive path. Physical books can be acquired cheaply from used-book markets and library sales, and digitization is often cheaper than licensing. The coalition's complaint suggests that destruction may not be incidental logistics but a strategic effort to prevent rivals from obtaining the same books once the original copies are gone. That distinction — between disposal as operational necessity and disposal as exclusionary conduct — will likely be central to any FTC review.

The immediate legal backdrop includes the August 2024 copyright lawsuit filed by author Andrea Bartz and other writers against Anthropic. That suit accused the company of acquiring, scanning, and discarding millions of print books to train models. In 2025, a judge ruled that Anthropic's use of legally purchased books to train its AI model Claude did not violate the asserted claims, according to CBS News. The coalition's letter does not rely solely on copyright theory; it asks the FTC to examine the destruction itself as an unfair method of competition. That shift from intellectual-property liability to market regulation could create new obligations for AI companies even where copyright defenses succeed.